Comparison of Type 2 Diabetes Pharmacotherapy Regimens Using Targeted Learning

The study will conduct head-to-head comparisons of type 2 diabetes mellitus (T2DM) treatment strategies using observational data from real-world clinical settings to assess cardiovascular disease (CVD) outcomes and other patient-centered outcomes in T2DM patients with moderate baseline CVD risk who are treated with each of these four classes of glucose-lowering medications known as SGLT2, GLP-1RA, DPP4, and SU. To mitigate bias concerns related to confounding and informative loss to follow-up, analyses will be based on modern causal inference methods combined with machine learning that emulate intention-to-treat (ITT) and per-protocol (PP) analyses of pragmatic randomized trials with active comparators to provide the most robust and precise estimates of relative and absolute effects we would expect in usual care settings.

Specific Aims and Hypotheses:

Aim 1. Compare the effect off SGLT21, GLP-1RA, DPP4, and SU on each study outcome in adults with T2DM when each type of medication is (a) initiated as second-line therapy after metformin, and (b) initiated as first-, second-, or third-line therapy, or after any history of glucose-lowering therapy independent of prior metformin use.

Aim 2. Compare the effect on each study outcome of earlier versus later initiation of SGLT2i, GLP-1RA, DPP4, SU as first-, or second-, or third-line therapy, or after any history of glucose-lowering therapy triggered by various changes in A1C, or CVD risk, or other patient characteristics.

Aim 3. Assess in each of the prior analyses whether the treatment effects on study outcomes vary across categories of baseline CVD risk and CVD event history, renal function, congestive heart failure status, age, sex and race/ethnicity, or other patient characteristics.